Oddbean new post about | logout
▲ ▼
Silberengel | 18 days ago (raw) | root | parent | reply | flag 
 Don't know if it's accurate, but that would be the cohort who got hit by ReplyGuy, right after joining.
▲ ▼
Eye Reply. | 18 days ago (raw) | root | parent | reply | flag 
 Aaah, yes. I forgot about the fool!
▲ ▼
tf | 18 days ago (raw) | root | parent | reply | flag 
 I have ~ no trust in the stats
There's no way these changes in variability are a real effect
https://i.nostr.build/tmzv3lIQtVY6Zbgz.jpg
▲ ▼
Ava | 18 days ago (raw) | root | parent | reply | flag 
 That chart is only in 5k increments. It checks out.
▲ ▼
tf | 18 days ago (raw) | root | parent | reply | flag 
 I mean the day-to-day variability
Prior to end of August you've got an eyeball variance of +/- 500 per day at the 95% confidence interval, superimposed on a linear trend

That started in January and lasted for 8 months

https://i.nostr.build/Ob9JcteETwEQPzM8.jpg

Then there's a step change to +/- 7,000 per day superimposed on I don't even know wtf to call that, which lasted about 2 months

Now it looks like we're back to +/- 500

I can't think of a reasonable model of human behaviour that would produce 10-fold step changes in variance